【作者】 靖鲲鹏；

【导师】 宋之杰；

【作者基本信息】 燕山大学 ， 管理科学与工程， 2013， 博士

【摘要】 由于社会的各个功能系统之间的关联度和依赖度不段增强，各类突发事件更容易演变为规模巨大、后果严重的非常规突发事件。非常规突发事件应急管理已成为多学科交叉的前沿课题。在大数据时代，通过信息可视化以支持管理决策和预测，成为应急管理领域的一个重要方向。如何从非常规突发事件的多元海量数据中，快速、准确地获取有价值的信息，并对其进行数据处理和可视化表达，实现多元信息融合，以支持应急处置中的智能决策过程，成为非常规突发事件应急管理研究领域的重要目标之一。本文以复杂系统科学、突发事件应急管理和信息可视化等作为理论基础，融合多学科理论，采用系统科学方法、管理科学、信息科学、数学、演绎归纳法、实证研究等方法，深入研究非常规突发事件应急管理中信息系统、信息流和数据特征等关键问题，构建非常规突发事件应急管理多元信息可视化系统，并研究多元信息可视化融合的理论、模型和方法，用于支持应急决策，具有重要的理论意义和现实意义。首先，本文根据中国应急管理系统结构模型，系统分析了我国应急管理总体框架、应急预案体系、应急管理系统结构，以及美国应急管理的核心系统—NIMS和NRP的框架结构、JFO的构成和部门结构，归纳了应急管理系统分层、递阶、多系统耦合并具有协调机制的本质特征。分析了应急管理中的信息流和数据管理技术在应急管理数据流中的应用，并重点分析了应急管理数据的海量、多源、异构、时变、时间敏感、难以共享和可信度不同等特征。界定了本文的研究对象—文本数据，比较了多维数据和多元数据，提出了降维和信息融合是多元文本数据处理的关键环节。在基于Web的应急管理数据总体框架基础上，构建了多元数据分层递阶图表示模型、分层递阶可视化融合和应急决策模型，设计了应急管理信息多元信息系统模型。其次，根据多元信息可视化系统的设计原则，分析了多元信息可视化系统的信息收集与处理、信息存储、多元图表示、信息融合、信息分析与利用、信息传播等基本功能；构建了多元信息可视化系统的逻辑结构模型，分析了其对应的四个层次和四个支持平台；分析了领导辅助决策系统、系统安全与维护平台两个支持结构。进而，设计了非常规突发事件应急管理多元信息可视化系统，并分析了该系统不同层级的子系统与应急管理预防、准备、反应和恢复等阶段之间的对应关系。再次，研究了基于雷达图的应急管理多元信息可视化融合方法。非常规突发事件应急管理系统作为一个复杂系统，在管理层面上是分层递阶的，既包含定性数据，也包含定量数据，呈现出多维、参数耦合的特征。提出了多层递阶参数耦合信息融合的一般模型。为了使该模型具有通用性，提出了输入数据的预处理方法：定量数据标准化方法——线性转换方法和归一化方法，定性信息预处理方法，定性信息与定量信息转换方法——简单语言概念生成、线性划分法、非线性划分法和二元对比插入法。根据特征选择和特征提取原理，建立了特征选择和特征提取的融合模型。基于雷达图原理，分析了高维数据分段特征融合与分层递阶的降维过程。提出了基于三角形面积和扇形面积的雷达图数据分析方法。最后，运用该方法进行实证分析。最后，研究了基于形式概念分析的应急管理多元信息可视化融合方法。根据应急管理多元数据分层递阶特征，分析了基于形式概念分析的多元数据可视化的可行性。以形式概念分析的原理与方法为基础，根据形式背景的分层递阶概念格生成算法，引入了形式背景的分层优化方法，提出了基于形式背景行列交换原理的属性偏序结构图表示方法。该方法可以对形式背景进行优化，绘制具有较显著层次关系结构的图形，实现属性聚类和多元信息分层递阶可视化。最后，运用该方法进行实证分析。更多还原

【Abstract】 The correlation and dependence are enhanced constantly among various socialfunction systems. All kinds of unexpected events are more likely to turn intounconventional emergencies with large scale and serious consequences. Unconventionalemergency management research has become a pivotal frontier and multidisciplinaryfield.One of major research goals in the field is how to obtain some valuable informationrapidly and accurately from data, information and knowledge included in unconventionalemergencies, and acquire multi-dimensional multi-variate (mdmv) information visuali-zation (InfoVis) expression through data processing and information fusion to support theintelligent decision-making process in emergency response.Complex system science, emergency management and information visualization weremade as the theoretical basis. The dissertation integrated multi-disciplinary theories andadopted the methods of system science, management science, information science,mathematics, deductive induction and empirical analysis. The research studied some keyissues like information system, information flow, data characteristics, etc in unconven-tional emergency management, and built unconventional emergency management visualmdmv information system, and proposed some models and methods about visual mdmvinformation fusion. It has important theoretical and practical significance.First of all, according to the emergency management system structure model of China,this article systematically analyzed the overall framework, the emergency pre-planssystem and the system structure of China’s emergency management, as well as the core ofAmerican emergency management system (the frame structure of NIMS and NRP, theconstitution and department structure of JFO). On this basis, it came to the conclusion thatthe emergency management system in essence is a layered, hierarchical, multi-systemcoupling system with a coordination mechanism. And then the paper analyzed theinformation flow in emergency management and the application of data managementtechnology in emergency management data flow, and the emergency management dadacharacteristics such as mass and multi-sources, heterogeneous formats, time-varying, time-sensitive, difficulty in sharing, and low credibility. Again the scope of this researchlies in the text data, and the comparison between the multidimensional data andmultivariate data was made, and the dimensionality reduction and information fusion wasproposed as a key link in the process of multiple text data processing. In addition, basedon the framework of a web-based emergency management data system, the mdmv datahierarchical graph model and the layered hierarchical visualization fusion supportingemergency decision-making model were developed. The emergency management mdmvinformation system model was also designed.Secondly, according to the design principles of mdmv InfoVis system, the basicfunctions of the system were analyzed such as information collection and processing,information storage, multiple graphs, information fusion, information analysis andutilization, and information dissemination. The study built a logic structure of the system,analyzed its four levels and four corresponding support platforms, along with two supportstructures-the auxiliary leader decision-making system and the system safety andmaintenance platform. And then, the unconventional emergency management visualmdmv information system was presented. The corresponding relationship was analyzedbetween the subsystems in the different system levels and four phases (prevention,preparation, response and recovery) in emergency management.Thirdly, the dissertation researched an emergency management visual mdmvinformation fusion method based on radar map. Unconventional emergency managementsystem as a complex system is layered on the management level. It contains bothqualitative and quantitative data, and presents a multidimensional and parameter couplingcharacteristic. A multi-level, hierarchical and parameter-coupling information fusionmodel was proposed. There is a need for input data preprocessing in order to make themodel general. Quantitative data standardization method was proposed. It contains lineartransformation method and normalized method. Qualitative information pretreatmentmethod was analyzed. Some conversion methods between qualitative information andquantitative information were developed such as simple language concept generationmethod, linear classification method, nonlinear insert division method and dual contrastinsertion method. According to the principle of feature selection and feature extraction, a fusion model of feature selection and feature extraction was established. Based on theprinciple of radar chart, this study analyzed the high-dimensional data segmentationfeature fusion and layered hierarchical dimension reduction process. In addition, somedata analysis methods based on triangle area and fan-shaped area of a radar chartrespectively were proposed. Finally, an empirical analysis was made by applying thesemethods.Finally, a study on visual mdmv information fusion method of emergencymanagement based on formal concept analysis (FCA) was developed. According to thelayered and hierarchical characteristic of mdmv data in emergency management, thereasarch analyzed the feasibility of mdmv data visualization based on FCA. Thedissertation proposed the attribute partial order structure diagram by applying the ranksexchange principle of formal context to optimize formal context, on the basis of theprinciple and method of formal concept analysis, and in accordance with the generationalgorithm of layered hierarchical concept lattice. The method can optimize the formalcontext, draw graphics with significant hierarchy structure, and realize attribute clusteringand hierarchical mdmv information visualization. Two cases illustrated the method.更多还原